Factorized Topic Models

Cheng Zhang, Carl Henrik Ek, Hedvig Kjellstrom

Decision: conferencePoster-iclr2013-workshop

Abstract: In this paper we present a new type of latent topic model, which exploits supervision to produce a factorized representation of the observed data. The structured parameterization separately encodes variance that is shared between classes from variance that is private to each class by the introduction of a new prior. The approach allows for a more efficient inference and provides an intuitive interpretation of the data in terms of an informative signal together with structured noise. The factorized representation is shown to enhance inference performance for both image and text classification.